Institute of Computer Science, Academy of Sciences of the Czech Republic
Pod vodárenskou vezí 2, 182 07 Prague 8, Czech Republic
Several recent thorough studies have confirmed a nonlinear component in the EEG dynamics, however, signatures of low-dimensional chaos were not found. These results pose the question about adequacy of applying so called chaotic measures (dimensions, Lyapunov exponents) in EEG analysis. It is shown that the chaotic measures applied to stochastic or even noisy chaotic data do not bring information not accessible by linear approaches such as spectral analysis. Moreover, even states of chaotic systems can be discernible by using an entropy rate computed from spectral densities. The applications of the chaotic measures do not seem to lead to a previously expected progress in the computerized EEG analysis, however, there are still ideas and tools developed in study of nonlinear (chaotic) systems, which could contribute to understanding the EEG dynamics and underlying brain processes as well as to improvement of clinical diagnostics. Perspectives for nonlinear dynamics in the computerized EEG analysis are seen in detection and characterization of nonlinearity in EEG dynamics and search for its physiological significance by comparing analyses of real EEG data and of signals generated by realistic models; in classifying complexity of the EEG signals by using entropy rates; or in detecting and characterizing synchronization of EEG signals recorded from different loci.